Christopher Miller

Welcome to the Puck Venture Capital and Beyond, the show that examines the changing landscape of our world. We'll have candid conversations with VCs, entrepreneurs, and thought leaders grappling with our current challenges and providing solutions to key problems we face as a nation. I'm Jim Beer, president of CMBG Advisors. This podcast brings change makers to the table to discover the inner workings behind their decision making strategies and ultimately how they got to where they are today. You know, you look back to the initial founders of some of the biggest chip firms.

I think of Bob Noyce, who founded Fairchild, semiconductor, and later founded Intel. And he was someone who, if you listen to what he was saying in the early sixties, he had a vision about things like home computers networked together. That was the Internet. Or personal portable communication devices. That was the smartphone.

So there were a small number of visionaries who really could look fifty years in the future and understand this will transform everything. Today, I'm sitting down with Chris Miller, a professor of international history at Tufts University, whose research centers on key shifts in technology, economics, and the changing geopolitical landscape. Miller discusses his newest book Chip Wars, which takes us on a fascinating journey into the development and proliferation of semiconductor chips dominating all aspects of our world. Let's get to it. Chris Miller, welcome to the show.

We are here to talk about your book, Chip War and the Fight for the World's Most Critical Technology. So before we get into the meats of this, what prompted you to start researching for the book? Well, I get interested in ships totally by chance, actually. I'm a historian of Russia by training. I started intending to write a book about the evolution of military technology during the Cold War, the Soviet US arms race.

Thought it was gonna be a story about nuclear weapons or missile systems or fighter jets and came to realize that actually the key factor in transforming the way militaries fight is the application of computing to military systems. And as I did that research, I learned that, in fact, the defense industrial base was the key driver of advances in semiconductors in the early fifties and late sixties and set the industry on the course it came to play. And so that's that was the initial impetus for my interest. But I I decided to read a book on the topic when I learned a different fact, which is that today, China spends as much money importing chips as spends importing oil, which helped me realize that the question of the chip industry wasn't just an interesting historical story about where did these companies come from, but it was also, I think, the central challenge in understanding how the world economy works today because there's no good that is treated as widely as semiconductors are and no product that's as central to the structure of globalization today. So before we get into kind of the military aspects of this and the kind of the new arms race, so to speak, you, in your book, talk about how chips have essentially created the modern world.

What do you mean by that? Well, most people don't realize it, but today, all of our lives, all of our economies, our society depends fundamentally on many thousand pieces of silicon, which we never see and which we're only dimly aware exist yet structure all aspects of our daily life. So I like to think about over the course of the day, the number of semiconductors I interact with. You wake up in the morning, turn off your alarm, there's a semiconductor in your alarm clock. Open your fridge, there's a chip inside of your fridge.

Turn into coffee maker, there's at least one chip, probably several inside of your coffee maker. Sit in your car, if it's a new car, it'll have a thousand chips on average inside of it. So even before you've opened your phone or turned on your computer, you've interacted with dozens, probably hundreds of semiconductors. And then the moment you do open your phone, you've got many dozens of semiconductors inside of your phone itself, dozens of semiconductors. But then you've also got all of the chips in the telecom network, in the data center that are processing, remembering, managing all of your data.

And so the typical person will interact with thousands and thousands of chips every day, even though we hardly ever see any of them. And what do these chips actually do? Well, there's a bunch of different types of chips, but to boil it down, chips are pieces of silicon in most cases, often the size of your fingernail with millions or billions of tiny circuits carved into them. These circuits flip on and off with help of a device called a transistor. And each one turning on or off creates either a one when it's on or a zero when it's off.

And all of the ones and zeros undergirding all computing come from these chips. And today, just to kinda put some numbers behind it, if you go to the Apple Store and buy a new iPhone, the primary chip, that iPhone will have roughly 15,000,000,000 transistors carved into it. Instead of fit 15,000,000,000 devices inside the size of your fingernail means that each one of them is roughly the size of a coronavirus. And so this is the most complex manufacturing that humans have ever undertaken. And in that regard, what is Moore's Law and how does that relate to these billions of transistors?

So today, we've got billions of transistors, but in the early sixties, when the first commercially available chip was, produced, they had four transistors on it. And so that progression from four to 15,000,000,000 illustrates the progress that's been made since then. And and Gordon Moore, who was one of the founders of Intel, realized in the mid sixties that the number of transistors per chip, which is roughly speaking its computing power, was doubling every year or two. And so he predicted at the time for the next ten years, from 1965 to 1975, that doubling rate would continue. It turns out that it would continue for over half a century.

And so the chip industry has had this exponential growth rate now for almost seventy years. And, you know, nothing else in human history has come remotely close to that rate of improvement. I like to think about, you know, what would aviation be like if airplanes flew twice as fast every two years? And the answer would be I actually calculated this. We'd be at roughly six times the speed of light right now flying across the Atlantic.

Physically impossible, but the chip industry has delivered that rate of growth thanks to Moore's Law. Yeah. It's interesting. My brother and I sometimes talk about technology and when you look at, as you said, cars and how they drive and airplanes how they fly and batteries, they haven't really had this unbelievable trajectory that chips have had. In your research in terms of doing that, is there something unique about chips that allow this law to apply and not in other areas?

Well, the key to Moore's law is that it wasn't only about the science and engineering of shrinking transistors smaller and smaller and smaller. It was also that as they shrunk, the average cost per transistor fell, which wasn't guaranteed, but companies found ways to make it possible. So that today, you can go on Amazon and buy a thumb drive with a billion transistors for $10.99. With a cost per transistor, that would have been inconceivable in the 1950s and 1970s or even the 1990s. And so that cost decline has made it possible to put chips into everything, into your fridge, into your microwave, into your coffee maker.

Whereas when chips were first invented, they were only possible to use in very, very high cost applications because they were super expensive. But now basic chips are so cheap, we use them without even thinking about them. Well and you talk about the complexity of the manufacturing process, and you talk about the size of these transistors being like the size of a coronavirus. How in the heck does machinery or technology make these transistors? I mean, it just it sounds impossible.

Yeah. You know, that that's the most, I think, incredible part of it. I when I started writing this book, I thought of computing is primarily about programming software, controlling ones and zeros, and I didn't realize the extent to which all computing depends on our ability to precision manufacture these chips out of silicon with basically perfect accuracy at coronavirus scale. And so if you're going to set up a cutting edge chip fab to make the chips that are inside of an iPhone or inside of a data center, you need to buy machine tools from just a handful of companies that have developed the capabilities to manipulate silicon on almost the atomic level. So for example, you need in cutting edge chip making deposition tools, which can deposit thin films of materials.

And the most sophisticated of these tools can put down thin films four atoms thick with basically perfect uniformity or etching tools that can etch tiny canyons into the silicon with roughly the same dimensions. And the the most complex tool is called the EUV, extreme ultraviolet lithography tool produced by, one, the company in The Netherlands, ASML, which has a 100% market share. And this machine is used to pattern the transistors onto a silicon wafer. And these machines require, among other components, inside of them, there's the flattest mirrors humans have ever made, bar none. There's one of the most powerful lasers ever deployed in a commercial device, and constantly, there's a explosion happening as tiny balls of tin are pulverized with a laser turning into a plasma at a temperature of roughly 40 times hotter than the surface of the sun.

And this is just one of the tools you need. And so when you look at these tools and they cost tens of millions, some cases, a $150,000,000 apiece. And that's that's what you need just to have the tooling necessary to start the manufacturing process of just one sunlight. Well, when you were on Scott Galloway, I mean, you talked about chips were like the new oil, but in your response, you talked about this concentration of their production. I mean, if this is the new oil and it's so important, how did we get to a situation where there was such extreme concentration?

Well, it really came as a surprise to a lot of people in in the industry and especially outside of the industry. It wasn't so long ago that there were actually a fair number of firms that could produce cutting edge chips. But as the engineering has gotten more complex and every year it gets more and more and more sophisticated and expensive, the number of companies that could keep up with just the changes in the science and engineering declined and the economies of scale grew and grew and grew. So today, if you wanna build a cutting edge chipmaiden facility, they'll cost you $20,000,000,000 of the most expensive factories in human history. And so, good luck raising that amount of money to start up a new company in that sphere.

And the whole industry is moving in that direction. And so Right. That has driven concentration across the supply chain. There's concentration in the manufacturing process. There's concentration in the manufacturing of the machine tools that are used in chip making.

There's concentration in the design of the ultra specialized software tools that are used in ship making. So the entire supply chain is in almost every step controlled by just a tiny number of firms that have unique capabilities to produce at the cutting edge. When I was reading your book, you know, I I start thinking about the ideas where truth is more interesting than fiction. And, you know, as a kid of James Bond, you know, spy novels and the whole thing about the KGB and the Russian stuff. As we talk about TSMC and and what's going on in Taiwan today and the concentration of chips, It is the most suspenseful kind of craziness that you can imagine when you think about where the world is going.

And so as we kind of unlock the players here, tell us a little bit about TSMC and and how there's so much importance to Taiwan, this little island that we are hearing more and more about in the news all the time because China is saying that they wanna reincorporate it in the mainland. This is starting to sound awfully, like I said, like a a spy novel. That's right. And TSMC is a company that most people have never heard of, even though everyone listening to this podcast relies on TSMC products every single day of their life. That TSMC produces around 90% of the most advanced types of processor chips.

So the chips in your phone, in your computer, in the data centers that store all your data, in the telecom infrastructure that makes it possible to use a mobile phone. TSMCs in all of that. Not just in it. They're the dominant producer. For companies like Apple, they produce all of Apple's key chips.

Qualcomm, Nvidia, they're the key producer. And so they've come to play this central role in all of the world's tech ecosystem. And they were just founded, not so long ago, thirty five years ago, by a entrepreneur named Morris Chang, who many people have have not heard of, but I think he ranks next to people like Steve Jobs and Bill Gates as among the key tech entrepreneurs of the last half century. He's had an extraordinary background, born in Mainland China, fled after the communist took power, enrolled at Harvard, where he was the only Chinese student in his class, got a job at Texas Instruments, and built up the chip industry in the early decades at TI, but was passed over for the CEO job. One of the biggest errors in twentieth century business history, not giving him the the CEO position.

But he left Texas Instruments and was looking for something new to do. He was already in his fifties, so pretty senior, but didn't wanna retire. And was looking at trends in the chip industry and realizing that there was going to be demand for more specialization in the future. And his key insight was the following. At the time that he founded TSMC, almost all chips were both designed and manufactured by the same company.

But he realized that as the production process was getting more complicated, there would be demands for specialization in both design and manufacturing separately. And so he decided he wanted to be sort of like what Gutenberg was for books. Gutenberg didn't write any books. He only printed them, but he printed them for lots of different authors. Morris Chang decided TSMC wouldn't design any chips.

They just manufacture them, but manufacture them for lots of different companies, for TSMC, for Apple, for Qualcomm, for Broadcom, for NVIDIA. And as a result, produces tremendous scale that has, let them drive down prices and reinvest in cutting edge technology. And so today, thanks to this business model innovation, they are undisputed the leading player when it comes to cutting edge processor chips, and the world simply couldn't live without them. And we're talking about this company that you talked about that, you know, where he raised this money thirty five years ago. And now when we hear about Taiwan and this important resource, we can just start thinking about Taiwan's relation with China.

Can you talk about this? I mean, how how does this relation with China work, and how is the approach that's being taken by China where they're being a little more aggressive? How does that play into this whole drama playing itself out? Yeah. It's a really complex set of interlocking dynamics.

On the one hand, you you've got the political dynamics which have existed before the first ships were even invented. So the nationalists fled to Taiwan in 1949. Since that point, the communists wanted to take control of Taiwan. And at that point, there weren't any chips. They hadn't yet been invented.

But overlaid on top of that, we've now got a ultra complex electronics and computing supply chain that can't live without Taiwan. And it's also increasingly interlinked with China. So if you take an iPhone, for example, know, if you look at the back of the iPhone, it'll say, designed in California, assembled in China, which is not untrue, but it's one of the sort of great lies of of the modern economy because the critical components of an iPhone, of course, are produced either in California or in China. They're manufactured in Taiwan. And so the entire world is just fundamentally dependent on chips made in Taiwan.

The US is, Europe is, Japan is, but also China is because the Chinese can't manufacture cutting edge ships. They import huge volumes of them from abroad and largely from Taiwan. So in some ways, that creates interdependence. Everyone would suffer immensely if Taiwan's chip making were knocked offline due to a a crisis or a war. But it also produces immense risk because there's simply no alternative to Taiwan's chip making capacity, and it would take years, literally years, to rebuild it.

And in the interim, if we were to lose access, we'd struggle to produce a smartphone anywhere in the world. Well and as somebody who wants his smartphone to work, obviously, that's important. But when you look at the military applications and kinda what led you down this line, how is it that The US and Japan, which for a long time, I mean, did control this market and were doing the manufacturing, how did companies like Intel and how did Taiwan actually become such a leader in this area and and essentially put us in this position? Or how did we allow ourselves stand up in this position? Well, in some ways, it was just pure fact that companies focusing on what they did best.

And so The US has always been a leader in chip design. Chip design is an immensely profitable business. US firms like Qualcomm, Nvidia, AMD have done extraordinarily well based on business models whereby they only design and trust, in most cases, TSMC to manufacture their chips. And so if you set aside the geopolitics, this has been a success story for American business. We've got these hugely successful firms, Nvidia's worth $1,000,000,000,000, and this has worked brilliantly.

But the fact that the manufacturing is so concentrated is something that I think most people didn't expect and didn't really realize until quite recently. There were two reasons they didn't realize it. First was that until around five years ago, five to seven years ago, Intel in The US was roughly comparable to TSMC in terms of manufacturing capabilities, and Intel was more at the center of key applications than it is today. So if you think of what Intel was founded around, it was found well, first founded actually built in memory chips. But more recently, it specialized in building chips for PCs and building CPUs for data centers, general processing chips for data centers.

But the PC market has been basically flat for a long time. No one excitedly upgrades their PC when a new processor comes out, they used to, not anymore. And in the data center, the types of chips that Intel has specialized in have become less prominent than GPU chips, which which are focused on applications, which we can talk about. Whereas TSMC has been more focused on growth markets, focused on a smartphone market, where Intel plays a tiny role, and also focused on the AI processor market, where NVIDIA largely turns to TSMC for manufacturing. So TSMC has overtaken Intel in terms of a lot of the growth markets, and that has made it possible for TSMC also to leap ahead in terms of its manufacturing technology.

And so seven years ago, you could say, well, we rely a lot on Taiwan, but the most advanced chips produced by Intel are produced in The US. But now as of right now, that's not true. TSMC has more advanced manufacturing capabilities than Intel. Intel wants to catch up. We'll see if they do.

But right now, TSMC is undeniably in the lead. The second thing is that as more and more applications have come to rely on chips, it matters more for the entire economy, whether we've got access. Twenty years ago, chips were mostly important for high-tech, for computers, for data centers, but today is everything. And the chip shortage of the pandemic illustrated the fact that car companies can't function without a secure supply of chips. And during the pandemic, just to give you one data point on this, a typical new car will have a thousand chips on average inside.

Some Some of them are pretty sophisticated. Others are really simple, like the ones that move your seat back and forth or your window up and down when you press the button. Car companies during the pandemic often just were missing one ship out of the thousand they needed to make their cars complete, and they couldn't sell cars. And so in aggregate, the car industry is estimated to have lost several $100,000,000,000 in revenue due to chip shortages during the pandemic. And that hammered home the extent to which now it's not just big tech firms that need chips.

It's everything that needs to secure supply chips. It's crazy. So when you go back and look at some of the issues in The Pacific and and World War two, there was an embargo against Japan and oil, and it led ultimately to Japan designing that it needed oil and there in low, we get bombed at Hawaii at Pearl Harbor. When you look at what's going on today, as I understand it, there's a bit of a quote embargo on China in terms of these chips and stuff. So question, am I describing right that we are actually trying to embargo the new oil to China, and that rhymed with what happened seventy years ago, so to speak.

And if that's true, how how well is it working? And if it is working well, what's China likely to to do about it? Or what are some of the possibilities? Because if this is as important as you say, then how big of effect is it gonna have on China? So I think there are some similarities and some differences with the the Japanese example.

So to put it in context, The US last October rolled out some new controls on technology transfer to China, focusing on two different things. First, banning the transfer of advanced GPU chips to China, which are used for AI training. And then second, banning the transfer of chip making tools to China so that China can't build up its own industry. And on the one hand, this certainly is an embargo when it comes to the types of hardware you need to efficiently train AI systems. And that's that was The US goal.

The US said, look. We think that AI is not just gonna be important for Chargebee T. It's gonna be important for defense and intelligence applications too. We know, and there's great open source research done on this, that all of the world's militaries rely on the same semiconductor supply chain to source all of the chips that they need. And until recently, you could find Chinese companies advertising their ability to procure American GPUs for Chinese military production.

So this was certainly happening. And all of the evidence from the AI community was showing that the amount of data used to train cutting edge AI systems was growing at an extraordinary rate every six to nine months, according to the best estimates, which means that the processing power required for training was gonna grow and grow and grow, which means that if you don't have access to the most advanced technology, you can train at best and efficiently, and at worst, not at all. So US said, why would we wanna give these very important pieces of hardware to a country that we see as a military adversary and that we know wants this hardware partly for civilian applications, but also partly for military applications. And so these controls were put in place. Now I think there's a difference with the Japanese example that's worth spending some time on because the Japanese case, if Japan had not gone to war to seize foreign oil fields, which is part of the cause of of the war, as you said, it would have been difficult to drive a car in Tokyo by the end of nineteen forty one or early nineteen forty two.

Whereas The US controls are only on the chips used for training advanced AI systems. In other words, they don't impact smartphones, don't impact PCs, don't impact autos, they don't impact almost any use except for AI training. And so 99% of the chips that China imports from abroad are not AI related and therefore not impacted. And so for most people in China, unless you're in the business of training AI systems, you're not gonna notice these controls this year or next year or the year after. Now over the long run, however, these controls are gonna matter because it's gonna make it much, much harder to cost effectively train AI systems in China, and that's their goal.

That's what The US is trying to do. Yeah. I mean, it's when you go back and you look at the arms' rates and the whole Manhattan Project and then how quickly the Soviets caught up and how we were surprised and the whole issue with Sputnik and everything. And when you look at this technology and what The US is trying to do and this kind of race to develop more and more sophisticated AI, and you take into consideration Moore's Law and everything else. If you had to guess and see where the puck is going, how material are these limitations right now?

Because if things continue to grow exponentially like they've been growing, and in fact, if The US maintains that lead, I would think that would be pretty intimidating to the rest of the world. Well, I think I think it is intimidating for the rest of the world. I think the question is, what are you gonna do about it? Because you can't just set up your own semiconductor supply chain from scratch. China is trying, and it's very, very, very difficult, expensive, and time consuming to do because of the extraordinary complexity involved.

And so most of the world's economies have decided they're gonna work within the confines of the existing supply chain. They're gonna be reliant on The US and Japan for certain types of tools and chemicals, on The Netherlands for lithography systems, for Taiwan when it comes to advanced manufacturing, and that they're comfortable with this because they have an alternative. China is the only country that can plausibly think it might have an alternative. And therefore, China already has spent the last decade trying to pour money into its chip industry and is gonna keep pouring money in, both because its domestic market is big enough to give it a shot, far from a guarantee, but a shot at building up its own self sufficient supply chain. And because its rivalry with The US and Taiwan and Japan means that the cost of being dependent are really quite large.

Well and if you game this out for a second, if it appeared, for instance, that The US was getting you know, when the West was getting very far ahead with this because of the chips in coming out of Taiwan, I mean, China has alternatives. It could try to blockade Taiwan or do some crazy things, which would obviously be horrific. I mean, to me, obviously, when you look at the chips act and everything that's going on, The US is trying, right, to get some of these plants built over here. And you talk in your book about how Intel stumbled. Why should we believe that Intel's not gonna stumble again?

Should we believe that we can actually pull this off? And essentially, if Taiwan were, for instance, to be blockaded, that in a few years, we actually would have the ability to continue to manufacture these AI chips? Well, I think if you look at what the the US government's doing with the Chips Act, they're not betting on any one company. They've been very clear. They're gonna give manufacturing incentives to a bunch of different companies, to Intel probably, to TSMC probably, to Samsung probably, as well as multiple other companies, which I I think is smart because, you know, you and I might bet on one company or another, but no one's certain.

And I think the government should be pretty modest in its ability to bet well. And so I think it's right to try to diversify the bets that are being made. And I think, you know, really rather than seeing that the government be the CHIPS Act is like a venture capital fund, which is, I think, not the right model, I think the better way to look at it is that in The US, we know that it's more expensive to run chip making facilities than in Asia for a variety of reasons, because Asian tax policy is better, because land and power costs are cheaper, because especially in China, there's huge subsidies available. And so the CHIPS Act is really not about helping specific firms, but it's about leveling the playing field in general so that The US doesn't find itself at a huge cost advantage and thereby companies are disincentivized from investing. And so I think that that is likely to lead to a lot more investing in investment in The US in chip making.

And you already see early signs of this. The amount of investment in new chip making facilities has actually jumped up dramatically over the last twelve to twenty four months. And I think that's evidence that in fact, you know, The US does have a pretty well developed chip making ecosystem. We have fewer chip manufacturers than Taiwan, but the machine tools are largely made in The US. The chips are largely designed in The US.

There's a pretty large chemical base in The US, better developed in Japan, but still pretty substantial. And so a lot of the ingredients you need from industry are already in The US. And so I think that's why companies are proving relatively willing to invest in The US even if they had it in the past because they're looking at the cost being more equalized thanks to ship back, plus The US having a very well developed ecosystem. And final point, if demand is in The US, what's the world's largest smartphone market? The US.

What's the world's largest data center market? The US. Your customers are ultimately here in The US as well. And that's that's a final and really important driver for chip companies when they decide to make decisions about where to invest. When you were doing, you know, the research for the book and learning all about chips and so forth, for people that grew up before chips were even a thing, is it was there any talk about what's gonna actually come after this?

In other words, is it gonna simply be that ships are the wave of the future in terms of these new weapons and technology and and guided munitions and everything? Or are there new types of switches and things that people are talking about that that even go beyond what we're doing with chips right now? You know, I think there's always been a question of in ten years, will the chip industry be able to deliver the similar performance growth that it had delivered the previous ten years? And if you look into the future, there's always uncertainty about what rate of growth it can deliver. Moore's law is a prediction, not a law.

It's proven correct for almost seventy years, but it's not guaranteed to work indefinitely. And so when we look into the future, we've got a very clear line of sight over the next, let's say, five, seven, maybe ten years to doubling performance on a biannual basis through them. Beyond that, nobody knows. Nobody ever knows how we're gonna be making improvements twenty or thirty years into the future. And so it's possible that quantum computing might become much more practically, applicable and might actually take up the mantle of kind of the driving the key advances in computing from silicon chips.

That's possible, far from guaranteed. There's lots of research into alternative computing paradigms, biological paradigms, for example. But so long as the chip industry keep Moore's Law alive, the promise of a doubling of performance within an existing paradigm with software tools, you know, with materials, you know, with manufacturing process, you know, is so, so attractive to the rest of the tech ecosystem that everyone will gladly choose that as opposed to anything that is new and more exotic and requires more adaptation. So if Moore's Law continues, and I think it will continue for at least a decade, we should assume we're gonna operate in a world defined by lots and lots of silicon chips. If I understand things correctly, the way we're building these factories and manufacturing these chips is essentially one of the most expensive, if not the most expensive thing that we've ever done.

But is it also correct that we are producing more of these items than anything else we ever manufactured? And how is that even possible? Yeah. That's right. We produce more transistors every year than there are cells in the human body.

The numbers are just mind boggling. Just for Steve, one more data point. Just for the primary chip on an iPhone, we produce a quintillion transistors a year. That's 10 to the eighteenth. That's just one of the chips on iPhone.

You have to learn, you know, new numbers to to study the industry because the the induced is followed by so many zeros. Yes. So there's nothing we've produced in this quantity. In fact, we produce more transistors than there are the combined quantity of all other types of goods in all of human history. There's just nothing else that comes remotely close.

So when you look at the AI revolution, because we're hearing more and more about AI all the time, and obviously, Chegg, GPT, and so forth, can you help us understand the relationship between chips and the AI revolution? I think the easiest way to put it is the following. You know, we know that there's there's three ingredients to AI. There's you need smart algorithms designed by smart people. You need data to train AI systems.

You need processing power to train the algorithm on the data. Now ask yourself this, which has gotten smarter at a faster rate over the past decade? Have we gotten people getting far smarter? Well, probably a bit smarter, but people don't get that much smarter. Have we gotten better data?

Well, maybe we've gotten more data from certain different applications, but probably not drastically better data. Or has processing power improved? Well, we know processing power has improved 16 times for the last decade because that's what Moore's Law promises. And so I think if you just put in those terms, whatever improvements there are in human education, they're not improving 16 times over the course of a decade. And so if you look at what has made the the current focus on AI possible, it's that we are now able to train AI systems with wild amounts of data.

And ChatGPT was basically trained by having or the OpenAI systems, which make possible ChatGPT, were basically trained by reading the Internet, reading Wikipedia, reading the entire corpus of books that exist, reading lots and lots of things written on Reddit. And that's only viable to do because advanced ships make that possible at a cost effective and time effective manner in a way that would have been simply inconceivable even just a decade ago. And so I think it's it's almost certainly the case that the primary driver advances in AI is actually not better algorithms or more finely selected data. It's it's just raw processing power. And if you look at what's expected over the next couple of years, NVIDIA, which is a company at the center of the production of these chips, has already promised multiple rounds, better chips over the next couple of years.

And so they're gonna keep this rate of growth going. And so, you know, I listen to people like Sam Altman, who runs OpenAI, say that he's might have to spend a $100,000,000,000 in capital expenditure to train all the AI systems he wants to train. Well, all of that money is gonna be spent on data center capacity. Data center capacity is just a bunch of servers in a room, and a server is just a bunch of very expensive chips inside of this metal rack. So almost a lot of money is going straight to buying semiconductors.

Well, as we're sitting here, because you spent a lot of time, and it's I think it's fascinating, talking about the history of a lot of the military, you know, weapons that came about and just the number of bombs that were dropped, for instance, on Vietnam that were wasted because they weren't smart and so forth. And you're talking about a $100,000,000,000 being spent here and all these crazy numbers. And then you think about the concern over chips and the defense application with these smart weapons and so forth. But is there any talk about how this AI is gonna change going forward from a defense perspective? Because again, we know it's gonna have guided missiles and, you know, smart weapons and so forth.

And I'm not saying we should terrify ourselves, but what other things are people worrying about and talking about in terms of the revolutionary effect of AI on the military, for instance? Well, I think if you look at the Russia Ukraine war, you already start to to see it. So, for example, the increasing use of drones in in the Russian Ukrainian war. We've had drones for a long time, but there's vastly more of them deployed now over the battlefields in Russia, Ukraine than previously. And these drones are increasingly semi autonomous, not completely autonomous, but they're doing more and more on their own.

And the Russians and Ukrainians are using simple drones. The Americans and the Chinese are investing in sophisticated drones. So I think autonomy, just like we're demanding that machines and civilian life become more and more autonomous, so too militaries are betting more and more heavily on autonomy. Second thing is communications and management of the electromagnetic spectrum. So if you go to the battlefields in Ukraine, open Google Maps and your phone, the GPS will often show you being hundreds of miles away because there's so much GPS spoofing.

And one of the things that the war has hammered home is that in modern conflict situations, both sides will be jamming the spectrum on a absolutely constant basis. And so, for managing your ability to communicate through a contested electromagnetic spectrum and also to jam your opponent to stop them from communicating and sensing. You need very sophisticated semiconductors, both semiconductors that are capable of sending signals through the air, but also capable of processing, understanding what your adversary is doing, jumping to free spectrum space, for example. And so there's more and more focus on using very sophisticated semiconductors for electromagnetic spectrum operations, electronic warfare, essentially, in military contexts. And the Ukrainians and Russians do that, but I think you have to assume that The US and the Chinese would be even better at it in a wartime situation.

And when you talk about autonomous drones, the chips are the biggest factor. They're like, when we're looking at autonomous cars coming down the pipe, you know, we know that the low orbit satellites, for instance, that Elon Musk is putting up, that's a component of this as well. But are the chips a big part of that? I mean, in terms of the ability to get self driving cars and these drones to be more autonomous? And I noticed in the news today, for instance, there was that story about how there there are autonomous cars driving around San Francisco right now, for instance.

How much new improvement needs to take place with chips and so forth to kind of take it to the next level? Yeah. You know, I think autonomous cars are actually a great analogy for thinking about autonomous drones in a military context. And so what are the chips you need for an autonomous car? Well, first off, you need tons of sensors on cars.

So cars now will really have a mix of LIDAR, radar, infrared, optical sensors. And so each of these sensors requires a, a chip to manage the sensor itself, and then a lot of processing in the car, plus memory, actually makes sense of the data. So you've got more and more chips on cars as the sensing demands get more sophisticated. But then as you train cars to drive autonomously, the chips that you're focused on are not only the chips on the car, important that they are. There are also chips in the data center that's training the car how to drive autonomously.

And so, you know, I think we're all on the early stages of understanding that actually data centers are where improvements to systems are gonna be increasingly hammered out in the future. Because your car's ability to drive more successfully than mine in an autonomous scenario is not primarily gonna be based on, does yours have better sensors than mine? The scope for difference there is small. It's who's got the best data running for the best algorithm. That's all happening in a data center.

So we've, we've got to increasingly think about not just the chips on the system, but the chips in the data center that are training that system. And then final point, you mentioned satellites. The communications network around an autonomous car is just as important because I think we we have to assume that, I mean, a, cars already have more and more communications capabilities embedded in them. And as we move further down the autonomy pathway, we'll be using more types of communications to undertake their activities, communicating with cars nearby, communicating with stoplights, communicating with satellites. And so drones are the same thing.

There's sophisticated chips in a drone, for sure, at least for more sophisticated types of drones, but it's actually in the data center and in the communications where you have even more demands for sophisticated semiconductors and processing power. And so militaries are having themselves to think about, it's not just what can I get in the nose cone of my missile, but it's how do I integrate that with my data center, with my sensors, with my communications platforms, and get an entire network that is hooked up, that has secure communications, reliable communications, and can transmit tons and tons of data around that? It's gotten just final anecdote. Militaries are now thinking about mobile data centers on the battlefield. So there are militaries that will bring data centers close to the front lines because they're so dependent on data.

They've got armored data centers to help them undertake the the communications and the sensing they need while fighting. So, I mean, it's mind boggling how fast this is going and how sophisticated it is. But when you talk about these data centers being relevant for autonomous cars and so forth, does that mean, like, what's going on in San Francisco? That there is a lot of work going into these data centers in San Francisco or wherever they're located to help these actual autonomous vehicles and that there is that symbiotic relationship between the two? Yeah.

That's right. And I think the way to look at almost every company today is is certainly an iceberg. You know, you you see the IT that the company operates itself, but then there's the iceberg below the water, which is some mix of AWS and Azure and Google Cloud, which are actually managing, in most cases, the majority of that company's data and helping provide the processing power. And so for companies that are doing really complex operations like autonomous driving, they've got this immense data center demands, which are usually provided at least in part by cloud computing providers. And so I think for for basically any segment of the economy that relies on automation of some sort, relies on AI for part of their operations, they've gotta think not just, you know, what's our in house tech look like, but also what's the data center infrastructure that we rely on.

So Chris, what's what's next for you? I mean, obviously, it's incredibly important to get the book out there for people to understand because again, I try to keep up on this stuff, but when I read your book, I was like, wow. This is such an amazing thorough analysis of kind of the most central issue of today. I mean, in the same way the oil embargo was a big deal that the whole world stopped and talked about, this whole drama over Taiwan is a completely different political situation when you take into account what you write about with these chips and so forth. So for our listeners, for Belts, we wanna do anything we can to help get this message out there.

But as you're seeing kind of where the world is going and you were surprised by this, because as you said, you were doing research in The Soviet Union and so forth in Russia. What's next for you in terms of where the world is going? What else is out there? Well, I'm really struck by as I watch the tech industry and the electronics industry shift, the extent to which the entire supply chains that go into, for example, the manufacturing of your smartphone are being transformed by this geopolitical competition. So today, you know, it's almost really the case that your smartphone is assembled in China using chips from Taiwan and Korea, but also the other components like the print circuit boards are almost certainly from China as well.

And now the entire world is is being split basically into two camps when it comes to tech and electronics. There's a a China supply chain and a non China supply chain, and every tech company is having to rejig the way they do business from Apple to HP to Dell. And that means opening up new facilities in Mexico and Vietnam and India. And it also means dividing their own employee base in half because they're trying to keep their employees in China while also developing a sort of walled off non Chinese operation. This is a huge challenge for basically every company from startups, all the way up to the biggest tech companies.

We do a lot of writing about inflation and budget deficits and, you know, one of the things that's so fascinating is that if you look at how much money the world has printed and how much debt we've taken, you know, gone into over the last ten or fifteen years, it's just a staggering amount. And globalization and this movement towards connectedness and low cost labor and everyone working together in that kumbaya moment when again Obama and the whole world got together in in the recession and kind of re monetize things. I mean, we are living in a different world right now that you help explain where China's got this unbelievable debt overhang in the real estate sector. There's all this discussion about do they print more money? How do they restart their economy?

What effect it's gonna have on the West and so forth? I mean, as you're saying all these supply chains, all of this is being reworked in real time. Do we have any idea where this is gonna end up? I mean, as you're looking at all these things going on and these changes, is it keep you up at night or do you think it's going, you know, smoothly? What are your thoughts there?

You know, I I think it's the process is gonna take a decade to play out, but, you know, to connect it to the inflation question, there's no doubt that if you are operating two supply chains in the world rather than one interconnected supply chain, you're gonna have more inefficiencies and therefore higher prices. And so I think we're now living through the reversal of what we've seen since the 1980s. That started a disinflationary period driven by a number of factors, but one of which was the emergence and the global market of a billion workers in China. And now we've got that being divided up. And there are replacements for China in some ways, but none of them are as good or as big.

So Vietnam, similar wages to China ten years ago, maybe fifteen years ago, but a tenth the size. And so it's already the case, even though there are a 100,000,000 people in Vietnam, it's already the case that companies say they struggle to find workers because the labor market is smaller. Now India's got the same population more or less as China's, but it itself is divided between different states and the infrastructures. It still needs lots of lots of work, lots of investment. And so there's going to be no perfect replacement for China.

There will be imperfect replacements, but the impact will be to add inefficiency to electronic supply chains. And we talk about this as like a new cold war, and we talk about peaceful coexistence and so forth. But the more you really think about this and these huge shifts that are taking place, it is very aggressive. I mean, we are we really are in the same way with the Soviet Union. We basically said, look, we have an arm strength and he who controls technology is going to ultimately win this.

I mean, we are it seems like we are really living in a different world now. What is China going to do? I mean, from their perspective, they have to see this as an existential threat. Yeah. I think arms race is right.

And, you know, people who spend their time thinking about the tech sector struggle to get their heads around that sort of framing. But if you look at what militaries in the region are doing, Japan last year announced it was going to double defense spending as a share of GDP. Well, that's the definition of an arms race. Right. Look at what Korea is doing, what Australia is doing.

And so I think if you look at what militaries are doing, they're operating under the assumption that they live in an arms race world. And I think that the tech sector was naive about this or just ignored it for a long time. And only the past couple of years has realized that whatever tech companies think, defense ministries think that tech companies are part of the arms race, every defense ministry is betting that their country's tech industry can provide critical components to next generation military systems. And so there's just no way of standing aside when all of the key governments see this as a existential challenge. Yeah.

And I guess president Biden had a summit with Japan and I think South Korea at Camp David recently and and you're seeing more and more coordination and talking about this. Do you think is there any movement in talking about pulling it back together where China and The US are are talking about cooperating on any of this stuff? Or is it really they're both in their own camps and pretty much locked down, so to speak? I think both sides talk about cooperation. Both sides talk about winding down tension.

But the reality is that the The US would be happy to wind down tension if China accepts that it's got a second place status in the technology industry. Right. And China talks about winding down tension if The US accepts that it's in charge of Taiwan. And both sides find the other one completely unacceptable, and that's why I think it's hard to imagine a actual reduction in tension for the foreseeable future. Well, let's try to end on a thoughtful and and or an optimistic view of things and stuff like that.

Where do you see the changes? I mean, in terms of what went on with vaccines, for instance, and the role of this technology to make people's lives better and so forth. Do chips play a role, for instance, in fighting diseases and new medicines and so forth? And can we look forward to optimism in terms of changes that are gonna take place in health care because of these chips? Well, actually, that's a great example of a way that chips have transformed an industry in a way that outsiders don't really realize, but in fact has been fundamental.

So if you talk to drug companies, one of the things they're most excited about is applying advanced AI techniques to drug discovery. Right. Because, research is extraordinarily time intensive, and you can have computers speed up your drug discovery cycles dramatically. And so that's a great example of a way that advances in semiconductors aren't just about smartphones or Apple or Amazon or tech firms. It's actually the entire economy that has benefited from the productivity enhancements that chips have made possible.

Well, and and again, we forget that Silicon Valley, the name itself. Right? And even Silicon Beach, they're calling down in Southern California. It all ties back to this and the importance in chips. And I I wonder, could anybody have imagined at the time that the name Silicon Valley came up in this whole world of silicon and sand for that matter could be such an instrumental part of our world today?

Well, there were very few people who imagined it. You know, you look back to the initial founders of some of the biggest chip firms. I think of Bob Noyce who founded Fairchild Semiconductor and later founded Intel. And he was someone who, if you listen to what he was saying in the early sixties, he had a vision of he would talk about things like home computers networked together. That was the Internet.

Or personal portable communication devices. That was the smartphone. So there were a small number of visionaries who really could look fifty years in the future and understand this will transform everything. Wow. Fascinating stuff and, you know, taking on something this complex and really making it readable, it's an amazing story that you tell and I I love the way you go through the history of it because it really is a spy novel.

Well, I'm glad you enjoyed it. It's a very kind word. So thank you for having me. Our pleasure. Appreciate it.

The POC Venture Capital Beyond is produced by CMVG Advisors. If you enjoyed the conversation today and haven't yet subscribed to our show, you can find us on iTunes, Apple Podcasts, and SoundCloud. Feel free to leave a review while you are there and maybe even a five star rating. Thanks for listening. We'll be back soon with a new episode.